Thermolysis of alkyl sulfoxides and derivatives: A comparison of experiment and theory

Article abstract of DOI:10.1021/jo0160625

Gas-phase activation data were obtained for model sulfoxide elimination reactions. The activation enthalpy for methyl 3-phenylpropyl sulfoxide is 32.9 ± 0.9 kcal/mol. Elimination by methyl vinyl sulfoxide to form acetylene has an enthalpic barrier of 41.6 ± 0.8 kcal/mol and that of 3-phenylpropyl methanesulfinate to form hydrocinnamaldehyde is 34.6 ± 0.6 kcal/mol. Calculations at the MP2/6-311+G(3df,2p)//MP2/6-31G(d,p) level for simplified models of these reactions provide barriers of 32.3, 40.3, and 32.7 kcal/mol, respectively. A series of other compounds are examined computationally, and it is shown that the substituent effects on the sulfoxide elimination reaction are much more straightforward to interpret if ΔH data are available in addition to the usually determined ΔH^?. The activation enthalpy of the reverse addition reaction is also subject to structural variation and can usually be rationalized on the basis of nucleophilicity of the sulfur or polarity matching between the sulfenic acid and olefin derivative.

Full text of DOI:10.1021/jo0160625

8722
J . Org. Chem. 2001, 66, 8722-8736
Th er m olysis of Alk yl Su lfoxid es a n d Der iva tives: A Com p a r ison of
Exp er im en t a n d Th eor y
J erry W. Cubbage, Yushen Guo, Ryan D. McCulla, and William S. J enks*
Department of Chemistry, Iowa State University, Ames, Iowa 50011-3111
wsjenks@iastate.edu
Received August 27, 2001
Gas-phase activation data were obtained for model sulfoxide elimination reactions. The activation
enthalpy for methyl 3-phenylpropyl sulfoxide is 32.9 ( 0.9 kcal/mol. Elimination by methyl vinyl
sulfoxide to form acetylene has an enthalpic barrier of 41.6 ( 0.8 kcal/mol and that of 3-phenylpropyl
methanesulfinate to form hydrocinnamaldehyde is 34.6 ( 0.6 kcal/mol. Calculations at the MP2/
6-311+G(3df,2p)//MP2/6-31G(d,p) level for simplified models of these reactions provide barriers of
32.3, 40.3, and 32.7 kcal/mol, respectively. A series of other compounds are examined computa-
tionally, and it is shown that the substituent effects on the sulfoxide elimination reaction are much
more straightforward to interpret if ∆H data are available in addition to the usually determined
∆Hq. The activation enthalpy of the reverse addition reaction is also subject to structural variation
and can usually be rationalized on the basis of nucleophilicity of the sulfur or polarity matching
between the sulfenic acid and olefin derivative.
In tr od u ction
The internal elimination of sulfoxides to form olefins
and sulfenic acids is well-established chemistry that has
seen considerable use in organic synthesis.1-3 Related
reactions include eliminations of thiosulfinates,4,5 N-
tosylsulfilimines6 (sulfoxide analogues in which the O
atom is replaced by an N-Ts group), selenoxides,7-9 and
at least in limited cases sulfones.10-13 In this paper, we
report an experimental and computational study on the
sulfinyl syn-elimination reaction that addresses struc-
tural effects on activation parameters and includes an
extensive set of molecules whose activation barriers are
predicted. It is argued that in many cases the enthalpic
barrier is controlled largely by the overall enthalpy
change of the reaction, but that between sets of closely
related reactions, the part of the barrier above the
endothermicity (i.e., the barrier to the reverse addition
reaction), can be related to conjugation in the transition
state, polarity matching between reacting partners, and
the acid/base properties of the molecules.
notation.14-18 This is done to emphasize the electronic
aspects of the reaction. For instance, the ylide notation
makes it clear that no S-O π-bond is broken. We argue
that the Ei reaction of sulfinyl derivatives is a direct
relative of the Cope elimination of amine oxides19-24 and
a more distant cousin of any pericyclic25 or pseudoperi-
cyclic26 reaction, such as the elimination reactions of
xanthates or carboxylic esters.27
Some review of the research in this area is in order.
Among the first lines of evidence for a concerted mech-
Throughout the paper, we use the ylide notation for
the sulfinyl group rather than the more traditional SdO
(14) Cioslowski, J .; Mixon, S. T. Inorg. Chem. 1993, 32, 3209-16.
(15) Dobado, J . A.; Martinez-Garcia, H.; Molina, J . M.; Sundberg,
M. R. J . Am. Chem. Soc. 1998, 120, 8461-71.
(16) Dobado, J . A.; Martinez-Garcia, H.; Molina, J . M.; Sundberg,
M. R. J . Am. Chem. Soc. 1999, 121, 3156-64.
(17) Cioslowski, J .; Surja´n, P. R. THEOCHEM 1992, 255, 9-33.
(18) Schmidt, M. W.; Yabushita, S.; Gordon, M. S. J . Phys. Chem.
1984, 88, 382-9.
(1) Trost, B. M. Chem. Rev. 1978, 78, 363-82.
(2) Trost, B. M. Acc. Chem. Res. 1978, 11, 453-61.
(3) Moghaddam, F. M.; Ghaffarzadeh, M. Tetrahedron Lett. 1996,
37, 1855-8.
(4) Block, E.; Ahmad, S.; J ain, M. K.; Crecely, R. W.; Apitz-Castro,
R.; Cruz, M. R. J . Am. Chem. Soc. 1984, 106, 8295-6.
(5) Block, E.; Ahmad, S.; Catalfamo, J . L.; J ain, M. K.; Apitz-Castro,
R. J . Am. Chem. Soc. 1986, 108, 7045-55.
(6) Oae, S.; Furukawa, N. Tetrahedron 1977, 33, 2359-67.
(7) Back, T. G. Organoselenium Chem. 1999, 7-33.
(8) Nishibayashi, Y.; Uemura, S. Top. Curr. Chem. 2000, 208, 201-
35.
(19) Cope, A. C.; Turnbull, E. R. Org. React. 1960, 317.
(20) Cope, A. C.; Foster, T. T.; Towle, P. H. J . Am. Chem. Soc. 1949,
71, 3929-34.
(21) Chiao, W.-B.; Saunders, W. H. J . Am. Chem. Soc. 1978, 100,
2802-5.
(9) Reich, H. J . Acc. Chem. Res. 1979, 12, 22-30.
(10) Cubbage, J . W.; Vos, B. W.; J enks, W. S. J . Am. Chem. Soc.
2000, 122, 4968-71.
(22) Bach, R. D.; Andrzejewski, D.; Dusold, L. R. J . Org. Chem. 1973,
38, 1742-3.
(23) Bach, R. D.; Braden, M. L. J . Org. Chem. 1991, 56, 7194-5.
(24) Bach, R. D.; Gonzalez, C.; Andres, J . L.; Schlegel, H. B. J . Org.
Chem. 1995, 60, 4653-6.
(11) Schmidt-Winkel, P.; Wudl, F. Macromolecules 1998, 31, 2911-
7.
(12) Kiran, E.; Gillham, J . K.; Gipstein, E. J . Appl. Polym. Sci. 1977,
21, 1159-76.
(13) Wellisch, E.; Gipstein, E.; Sweeting, O. J . J . Appl. Polym. Sci.
1964, 8, 1623-31.
(25) Woodward, R. B.; Hoffman, R. The Conservation of Orbital
Symmetry; VCH Publishers: Weinheim, 1970.
(26) Ross, J . A.; Seiders, R. P.; Lemal, D. M. J . Am. Chem. Soc. 1976,
98, 4325-7.
10.1021/jo0160625 CCC: $20.00 © 2001 American Chemical Society
Published on Web 11/28/2001

Thermolysis of Alkyl Sulfoxides and Derivatives
J . Org. Chem., Vol. 66, No. 26, 2001 8723
Sch em e 137,38
anism was the result that stereochemical integrity is
usually maintained during sulfoxide elimination reac-
tions. A particularly important example is the pyrolysis
of trans-4-methylcyclohexyl p-tolyl sulfoxides, in which
the two enantiomers of the sulfoxide lead largely to
opposite enantiomers of 4-methylcyclohexene.28
Pyrolysis of appropriately substituted sulfoxides in
either the gas or solution phase usually displays first-
order kinetics and is usually not affected by radical
inhibitors, again consistent with a concerted mechanism.
However, as far back as 1960, Kingsbury and Cram
suggested there might exist a competing homolytic mech-
anism for 1,2-diphenylpropyl phenyl sulfoxides.29 This
was based on curved Eyring plots over the fairly narrow
range of 70-110 °C and the loss of stereospecificity in
the higher temperature regime. There has not been
convincing evidence for a radical mechanism involving
any other substrate, but this may depend on the presence
of at least the R-phenyl group, which should stabilize the
putative benzyl-type radical.
cycloheptyl, cyclohexyl, and cyclopentyl phenyl sulfoxide
were 120:1:25, respectively. The slowness of the cyclo-
hexyl system was attributed to an awkward flattening
of the ring required to achieve the proper transition state
geometry. Similar results were later obtained by Yoshimu-
ra and co-workers.36
Phenyl sulfoxides are known from the synthetic lit-
erature to eliminate at lower temperatures than the
corresponding methyl sulfoxides. Emerson and Korniski
correlated substituent effects with the rate of reaction
for a series of substituted-phenyl propyl sulfoxides.37 At
any given temperature, a Hammett plot gave a modestly
positive F, meaning that electron-withdrawing groups
accelerate the reaction. While the activation enthalpies
(25.3 ( 2.6 to 27.9 ( 2.1 kcal/mol) follow this trend, the
range is modest and they are only marginally outside of
experimental error from one another. The ∆Sq values
(-11.5 ( 6.7 to -16.0 ( 8.2 cal/mol‚K) are similar in that
the values are not widely beyond experimental uncer-
tainty from one another. These results were rationalized
in terms of negative charge buildup on sulfur in the
transition state. A similar positive F value was found for
the pyrolysis of substituted-phenyl tert-butyl sulfoxides
by Shelton and Davis.33
In a related study, Yoshimura and co-workers studied
the decomposition of substituted 1-phenylethyl phenyl
sulfoxides in dioxane.38 Hammett plots showed a positive
F for the X position in agreement with Emerson and
V-shaped Hammett plots for the Y-substituents. Scheme
1 illustrates the Emerson and Yoshimura structures.
Activation data obtained over a 20 °C range were also
generally consistent with Emerson’s results. In addition,
they found a kinetic isotope effect of kH/kD ) 4-6 in the
range of 80-100 °C. The authors suggested that electron-
withdrawing groups on Y promoted a nearly synchronous
reaction, but in other instances, an E1-like transition
state was in force.
The effects of various structural parameters on the
kinetics of the sulfoxide elimination have been investi-
gated over the years. Emerson and co-workers examined
the effect of alkyl branching and found that the overall
rate for the formation of alkene was greater for the more
branched carbon substituent.30 For example, sec-butyl
ethyl sulfoxide eliminated faster than n-butyl ethyl
sulfoxide. Gas-phase activation enthalpies were near 30
kcal/mol, and the activation entropies were between -3.6
and -17 cal/K‚mol. Walling and Bollyky investigated the
pyrolysis of methyl 3-phenylpropyl sulfoxide in diglyme
and got a similar activation energy of 31.6 ( 3 kcal/mol.31
The elimination of isobutylene from di-tert-butyl sul-
foxide occurs faster than related but less sterically
demanding systems. The barrier is below 30 kcal/mol,
though some curvature in the data was observed.32,33
Kwart attributed the lower barrier to a “corset effect” in
which the starting material is destabilized by sterics
more than is the transition state and the barrier for
proton transfer is narrowed.34 Kice and Campbell studied
the effect of ring size on the rate of pyrolysis of cycloalkyl
phenyl sulfoxides.35 At 130 °C, the relative rates of
Despite the Hammett plot results in which the sign of
F is opposite for the X substituent’s effects on the
elimination reaction and on sulfoxide basicity,39 some
authors have considered structural effects on the Ei
reaction in terms of acid/base properties related to the
proton abstraction by the oxygen atom. The acidity of the
abstracted proton, unless controlled for with other ad-
equate structural detail (see the Discussion), does not
appear to have a consistently large effect. Elimination
of the illustrated R- and â-carbethoxy-substituted sul-
foxides affords the same products,40 but the abstracted
proton is substantially more acidic in the â-structure. At
(27) Birney, D. M.; Xu, S.; Ham, S. Angew. Chem., Int. Ed. 1999,
38, 189-93.
(28) Goldberg, S., I.; Sahli, M., S. J . Org. Chem. 1967, 32, 205-
2062.
(29) Kingsbury, C. A.; Cram, D. J . J . Am. Chem. Soc. 1960, 82,
1810-9.
(30) Emerson, W. W.; Craig, A. P.; Potts, I., J r. W. J . Org. Chem.
1967, 32, 2.
(31) Walling, C.; Bollyky, L. J . Org. Chem. 1964, 29, 2699-701.
(32) J anssen, J . W. A. M.; Kwart, H. J . Org. Chem. 1977, 42, 1530-
3.
(36) Yoshimura, T.; Sekioka, T.; Tsukurimichi, E.; Simisaki, C.;
Hasegawa, K. Nippon Kagaku Kaishi 1992, 817-23.
(37) Emerson, D. W.; Korniski, T. J . J . Org. Chem. 1969, 34, 4115-
8.
(33) Shelton, J . R.; Davis, K. E. Int. J . Sulfur Chem. 1973, 8, 197-
(38) Yoshimura, T.; Tsukurimichi, E.; Iizuka, Y.; Mizuno, H.; Isaji,
H.; Shimasaki, C. Bull. Chem. Soc. J pn. 1989, 62, 1891-9.
(39) Andersen, K. K.; Edmonds, W. H.; Basiotti, J . B.; Strecker, R.
A. J . Org. Chem. 1966, 31, 2859-62.
204.
(34) Maier, G.; Pfriem, S.; Schafer, U.; Matusch, R. Angew. Chem.,
Int. Ed. Engl. 1978, 17, 520-1.
(35) Kice, J . L.; Campbell, J . D. J . Org. Chem. 1967, 32, 1631-3.
(40) Crich, D.; Lim, B. L. J . Chem. Res., Synop. 1987, 11, 353.

8724 J . Org. Chem., Vol. 66, No. 26, 2001
Cubbage et al.
Ta ble 1. Cop e Elim in a tion Rea ction Activa tion En er gies
(Ea ) a t Va r iou s Levels of Th eor y for th e Ei Rea ction of
Eth yla m in e Oxid e
typically between CCSD(T) and B3LYP. A similar trend
was observed in the sulfinyl elimination reaction calcu-
lated by J ursic:42 BLYP/6-31G(d) < MP2/6-31G(d) <
HF/6-31G(d).
level of theory/basis set
Ea (kcal/mol)
Below, we present gas phase activation barriers for the
thermolysis of several sulfoxides and related derivatives.
A computational protocol is established that closely
matches the experimental values. With these data in
hand, extensions of the calculations to a number of
molecules not studied experimentally is reasonable, and
we are able to survey the structural effects on the
sulfoxide elimination and analyze them in terms of both
the elimination and the reverse addition reactions.
HF/6-31G(d,p)
43.3
26.3
24.3
28.8
MP2/6-311++G(d,p)
B3LYP/6-311++G(d,p)
CCSD(T)(full)/6-311++G(d,p)a
a
Calculated at the MP4(SDQ)/6-311G(d,p) geometry.
70 °C, the rate of elimination of the â-carbethoxy pyridyl
sulfoxide is higher by a factor of 1.5. Tsukurimichi and
co-workers, however, found that, at 90 °C, the rate of
decomposition of 2-carbomethoxyethyl phenyl sulfoxide
was only about one-third that of 1-carbomethoxyethyl
phenyl sulfoxide.41
Resu lts
Ap p r oa ch . Experimental and computational data
were sought for a variety of sulfoxides and related
derivatives in order to establish a reasonable computa-
tional protocol. The experimental systems needed to
provide clean reactions and products that were easy to
detect. Most sulfoxides were designed to give allylbenzene
or an analogous product. On the other hand, for the
calculations, the important electronic parts of the mol-
ecule must be retained, but any “excess” parts must be
stripped off to make the calculations practical. As a
result, the molecules calculated and run experimentally
were not, in general, the same, though they were found
to model each other very well. The molecules used in the
study are shown in Chart 1. Where experimental data
were collected, a related compound appears with an “e”
appended to the number to denote the actual molecule
used in the experimental studies. The proton that is
transferred during the reaction is illustrated explicitly
for each compound.
We are aware of only a single computational study of
the syn-elimination of sulfoxides.42 J ursic compared the
eliminations of a corresponding amine oxide, sulfoxide,
and phosphine oxide and showed that the amine oxide
elimination occurred with the lowest activation enthalpy
regardless of level of theory applied, with the sulfoxide
being the intermediate case.
The related Cope elimination24,42,43 of amine oxides has
been subject to computational study and is instructive.
Reasonable agreement has been achieved between acti-
vation barriers for model compounds ethylamine oxide
and 3-butenylamine oxide, calculated at the MP2 level
of theory, and experimental data for related sub-
strates.21,22,24,43
Once a reasonable computational protocol was estab-
lished, the method was extended to a series of groups of
compounds. Each of the groups contains compounds
around which a particular structural question could be
addressed. Though only illustrated in group 1, ethyl
methyl sulfoxide (1) is taken as the benchmark against
which the other compounds are judged.
The rationale for these various groups can be stated
briefly. Group 1 was chosen to investigate the effect of
substituents conjugated to the sulfinyl group. Group 2
stresses the electronegativity of the same “non reactive”
substituent, while group 3 addresses the “corset effect”
alluded to in the Introduction. Group 4 considers the
effect of having the abstracted proton on an sp2 carbon
and the homologation of the reaction. Group 5 was
designed to examine the effect of acidity of the proton or
polarity of the reverse reaction with models of syntheti-
cally relevant systems. Group 6 provides data on endo-
and exocyclic eliminations as a function of olefin ring size.
Group 7 is designed to couple the effect of formation of
double bonds with a heteroatom with another look at
acidity. The compounds in group 8 were chosen to
compare to literature results in which interactions be-
tween bond dipoles had been invoked to explain selectiv-
ity.
The barriers calculated by Tronchet and Komaromi at
various levels of theory are shown in Table 1.43 Since a
good-quality gas-phase experimental value is not avail-
able for the parent case, we must accept the CCSD(T)-
(full)/6-311++G(d,p) value as the standard for compari-
son to the other methods. It has long been known that
HF barriers are too high for this type of reaction.44
Despite its current popularity in the literature, B3LYP
has its shortcomings, and one that is well-documented if
not widely appreciated is in cases where proton transfers
are involved.45-47 Despite differences in the basis set in
Table 1, it is a consistent observation that barriers are
too low when calculated by B3LYP. The MP2 values are
(41) Tsukurimichi, E.; Yoshimura, T.; Yoshizawa, M.; Itakura, H.;
Shimasaki, C. Phosphorus, Sulfur Silicon 1989, 1989, 113-20.
(42) J ursic, B. S. THEOCHEM 1997, 389, 257-63.
(43) Komaromi, I.; Tronchet, J . M. J . J . Phys. Chem. 1997, 101,
3554-60.
(45) Barone, V.; Orlandini, L.; Adamo, C. Chem. Phys. Lett. 1994,
231, 295-300.
(46) Barone, V.; Orlandini, L.; Adamo, C. Int. J . Quantum Chem.
1995, 56, 697-705.
(44) Hehre, W. J .; Radom, L.; Schleyer, P. v. R.; Pople, J . Ab Initio
Molecular Orbital Theory; J ohn Wiley & Sons: New York, 1986.
(47) J ohnson, B. G.; Gonzales, C. A.; Gill, P. M. W.; Pople, J . A.
Chem. Phys. Lett. 1994, 221, 100-8.

Thermolysis of Alkyl Sulfoxides and Derivatives
J . Org. Chem., Vol. 66, No. 26, 2001 8725
Ch a r t 1
Exp er im en ta l Resu lts. The preparation of the sul-
foxides was generally routine, and details are provided
in the Experimental Section and Supporting Information.
Several attempts were made to prepare sulfoxide 4e from
the corresponding sulfide without success.
Thermolyses were carried out in a pulsed stirred flow
apparatus that uses a furnace area followed by an on-
line GC to quantify starting materials and products.48
Sulfoxide samples were injected as concentrated solutions
in acetonitrile, and He was the carrier gas. Activation
data are obtained by running the furnace at several
different temperatures, but the range is limited by the
need to quantify both starting materials and product.
Useful data were only obtainable when activation en-
thalpies were about 30 kcal/mol or greater. The utility
of the system is also limited by the need for the reaction
to be very clean. A typical set of data is shown in Figure
1, and the activation data are given in Table 2. Plots for
the other compounds are in the Supporting Information.
All data reported here are for reactions in which only the
expected olefinic product was observed. Formation of the
sulfenic acids is inferred, since none of them survived GC
analysis. However, their formation is well established by
previous studies in solution. Though sulfenic acids are
not generally stable in solution, they can be isolated as
their anhydrides (thiosulfinates) or trapped with acti-
vated olefins.49,50 There is also evidence, by means of
F igu r e 1. Arrhenius and Eyring plots for 3-phenylpropyl
methanesulfinate (270-340 °C) elimination to give 3-phenyl-
propanal. Data points are from three runs at each tempera-
ture.
deuterium incorporation, of sulfenic acid trapping by
unactivated olefins in unimolecular cases, as in certain
penicillin derivatives.51
(49) Structural Chemistry; Barrett, G. C., Ed.; J ohn Wiley and Sons
Ltd.: New York, 1990.
(50) Hogg, D. R. In The Chemistry of Sulphenic Acids and their
Derivatives; Patai, S., Ed.; J ohn Wiley & Sons Ltd.: New York, 1990;
p 361-402.
(48) Baldwin, A. C.; Davidson, I. M. T.; Howard, A. V. J . Chem. Soc.
1975, 71, 972-9.

8726 J . Org. Chem., Vol. 66, No. 26, 2001
Cubbage et al.
Ta ble 2. Activa tion P a r a m eter s for th e P u lsed
Stir r ed -F low Th er m olysis of 1e, 2e, 11e, a n d 26ea
set (6-31G(d,p)) gave answers in the right ballpark,
compared to experiment.
sulfoxide
1e
2e
11e
26e
The salient point to be gleaned from Table 4 is that
while ∆Hq
energies appear to have converged with
elim
T range (°C)
240-300
225-270
340-400
270-340
log A (s-1
)
12.5 ( 0.3 12.1 ( 0.8 13.1 ( 0.3 10.9 ( 0.2
34.0 ( 0.9 30.8 ( 0.8 42.9 ( 0.8 35.7 ( 0.6
respect to basis set, the convergence of ∆Hrxn is less clear.
Among the Pople basis sets, addition of the diffuse
functions is not a significant factor for ∆Hqelim, nor is the
doubling vs tripling in the 31 or 311 split valence.
However, to have a balanced basis set,53,54 we felt the
need to have the triple split valence. Thus, adoption of
6-311+G(3df,2p) seemed like a reasonable approach,
especially given the ambiguity over the ∆Hrxn value.
Going to anything bigger would have been difficult for
many of the relatively large systems (e.g., 3), but as
indicated below, good results were obtained for the
systems in which both computational and experimental
models were available. This and related10 success suggest
that this basis set gives data that are at least close to
the infinite basis set limit for ∆Hqelim. The Dunning basis
sets were considered but rejected. Those sets that would
have sufficient polarization functions (e.g., aug-cc-PVTV)
would have so many basis functions overall that the
computational resource needs would be prohibitive for
all but the smaller systems.
Coupled with the choice of basis set was the choice of
level of theory. We seriously considered three levels of
theory: MP2, B3LYP, and MCSCF with perturbation
theory correction (i.e., MRMP2). A modest test set that
included most of the experimental compounds was used
for evaluation. Activation energies from the MCSCF
calculations (vide infra) were very similar to the HF
values, so the multireference perturbation theory ap-
proach was not pursued further. Because the CCSD(T)
calculations, which were being taken as the standard
with regard to level of theory, would be impractical with
the chosen basis set, the level-of-theory calculations were
done with the 6-31G(d,p) basis. Table 4 shows the
calculated values for several representative sulfoxides.
Ea (kcal/mol)
∆Hq (kcal/mol) 32.9 ( 0.9 29.8 ( 0.8 41.6 ( 0.8 34.6 ( 0.6
∆Sq (cal/mol K) -4.5 ( 0.8 -6.5 ( 0.8 -2.1 ( 1.2 -12.1 ( 1.0
a
Errors are expressed as two standard deviations of the least-
squares fit.
Ta ble 3. Ba sis Set Eva lu a tion a t MP 2 Level on th e
Elim in a tion Rea ction for Su lfoxid e 1a
∆Hq b
(kcal/mol)
∆Hrxn
(kcal/mol)
b
basis functions
basis set
on 1
6-31G(d,p)
119
139
147
149
199
223
253
274
151
179
211
231
235
255
305
119
201
299
479
32.9
33.4
33.2
32.7
33.9
33.8
36.3
36.3
29.8
31.2
33.0
33.1
32.7
33.3
36.3
25.1
27.4
31.7
32.4
33
16.6
16.5
16.7
19.9
23.8
23.9
27.3
26.3
18.1
15.6
6-31+G(d,p)
6-31++G(d,p)
6-31G(2d,p)
6-31G(2df,p)
6-31G(2df,2p)
6-31G(3df,2p)
6-31+G(3df,2p)
6-311G(d,p)
6-311++G(d,p)
6-311G(3d,p)
6-311G(2df,p)
6-311G(3d,2p)
6-311G(2df,2p)
6-311+G(3df,2p)
cc-PVDZ
23.9
23.4
26.3
aug-cc-PVDZ
cc-PVTZ
aug-cc-PVTZ
Experiment
a
Single-point calculations at various basis sets on the optimized
MP2/6-31G(d,p) geometry. b No ZPEs are included. ZPE corrections
lower ∆Hq by about 3.0 kcal/mol and ∆Hrxn by about 3.7 kcal/mol.
Some compounds were run on the system, but proper
activation data could not be obtained. Sulfoxide 3e was
on the margin of having too low of an activation barrier,
and the temperatures required to get it through the GC
made collecting good data impossible. Sulfinamide 6e
underwent the elimination reaction, but chromatographic
conditions could not be found that allowed for accurate
integration of the peak area of the starting material. A
related tert-butylsulfinamide did not give clean chemis-
try. Similarly, 11e2 gave multiple products. However, the
elimination to find the acetylene was an important
experimental result to get. As a result, 11e (same as 11)
was prepared. To give a detection window in the GC
traces that avoided solvent, mesitylene was used as the
solvent rather than acetonitrile.
As with the Cope elimination, HF activation enthalpies
are too high and B3LYP activation enthalpies are too low.
The MP2 and CCSD(T) values are quite comparable in
most cases and in the general vicinity of the experimental
values. It remains unclear why there is such a large
difference between the MP2 and CCSD(T) calculated
∆H for 10. Nonetheless, general agreement between
CCSD(T) and B3LYP is worse. A final test of B3LYP,
with the larger basis set, was conducted, and the results
are shown in Table 5. The ∆Hq
values remain too
elim
small. Thus, MP2/6-311+G(3df,2p)//MP2/6-31G(d,p) was
adopted as the general method.
Com p u ta tion a l Resu lts: Meth od ology. The work
of Turecek52 and others (including some of our own
unpublished work) suggests that fairly large basis sets
are required to correctly model sulfoxide energies relative
to isomeric sulfenic esters (i.e., R-S-O-R′). Increasing
the size of the basis set, in particular adding polarization
functions on sulfur, stabilizes the sulfoxide relative to the
sulfenate.
Two sulfoxides (1 and 11, along with the respective
transition states and products) were optimized at the
MP2/6-311+G(3df,2p) level to see the effects on the
energetics and geometries of re-optimizing rather than
using MP2/6-31G(d,p) geometries for single-point calcula-
tions. The elimination activation enthalpies increased by
0.3 and 0.7 kcal/mol, respectively. The reaction enthalpies
changed by 0.6 and 1.0 kcal/mol. This was deemed too
insignificant to justify optimization in every case. As a
result, the MP2 calculations cited below are MP2/6-
311+G(3df,2p)//MP2/6-31G(d,p). Intrinsic reaction coor-
The effect of the basis set size on ∆Hq
and ∆Hrxn for
elim
the Ei reaction of 1 are shown in Table 3 with both Pople-
type and Dunning’s correlation consistent basis sets. The
MP2 level of theory was chosen for basis set evaluation
after some initial experimentation with a modest basis
(53) Gordon, M. S.; Truhlar, D. G. J . Am. Chem. Soc. 1986, 108,
5412-9.
(51) Cooper, R. D. G. J . Am. Chem. Soc. 1970, 92, 5010-1.
(52) Turecek, F. J . Phys. Chem. A 1998, 102, 4703-13.
(54) Gordon, M. S.; Truhlar, D. G. Int. J . Quantum Chem. 1987,
31, 81-90.

Thermolysis of Alkyl Sulfoxides and Derivatives
Ta ble 4. Activa tion En th a lp ies a n d Hea ts of Rea ction for Su lfoxid es 1, 2-4, 10, a n d 11a
10
J . Org. Chem., Vol. 66, No. 26, 2001 8727
1
2
3
4
11
∆Hq
∆H
∆Hq
∆H
∆Hq
∆H
∆Hq
∆H
∆Hq
∆H
∆Hq
∆H
HF
MP2
CCSD(T)
B3LYP
exptb
44.0
28.9
29.6
23.5
33
0.4
12.8
8.5
43.5
28.6
29.3
22.5
30
-2.1
9.6
7.8
42.7
28.5
29.2
22.3
0.5
14.5
9.5
41.2
27.4
27.6
21.9
-3.5
9.8
5.0
56.0
53.7
49.4
41.4
-5.2
16.6
6.3
50.1
37.9
39.4
34.4
42
15.0
20.4
19.5
22.9
9.0
8.4
8.5
6.8
6.4
a
All values are in kcal/mol and represent fully optimized structures at the specified level of theory with the 6-31G(d,p) basis set with
ZPE corrections, save the CCSD(T) calculations, done at the MP2 geometry and corrected with the MP2 ZPEs. Experimental ∆Hq are
for compounds 1e, 2e, and 11 ) 11e.
b
Ta ble 5. Activa tion Ba r r ier sa a t
MP 2/6-31+G(3d f,2p )//MP 2/6-31G(d ,p ) a n d
B3LYP /6-31G+(3d f,2p )//B3LYP /6-31G(d ,p )
1
2
3
4
11
MP2
32.3
29.2
33
30.4
26.9
30
30.6
26.6
30.7
27.9
40.3
40.0
42
B3LYP
exptb
a
b
Values are in kcal/mol and include ZPE. For the correspond-
ing “e” compounds.
F igu r e 3. Geometry of 1, 11, and the respective transition
states 1TS and 11TS. All bond distances are shown in Å. All
bond distances and angles are shown in the following order:
HF/6-31G(d,p), (MP2/6-31G(d,p)), [MP2/6-311+G(3df,2p)], and
{B3LYP/6-31G(d,p)}.
F igu r e 2. IRC calculation for the elimination reaction of 1
at MP2/6-31G(d,p).
dinate (IRC) calculations55 were carried out on com-
pounds 1, 10, 11, and 26-29. The transition states
clearly connected to the starting materials and products.
The IRC for 1 is shown as Figure 2. The others are in
the Supporting Information.
Geometries for ethyl methyl sulfoxide (1), its transition
state (1TS), and methyl vinyl sulfoxide (11) and its
transition state (11TS) are shown in Figure 3. They are
representative. The geometries, as expected, are all fairly
similar with respect to method, with the exception of the
HF geometry of 11TS, where the C-H-O parameters
are considerably out of agreement with any of the
correlated models.56 The result that HF/6-31G(d,p) bond
lengths for 1 and 11 are very close to experiment57 would
appear to be due to fortuitous cancellation of errors from
basis set (too long) and level of theory (too short). With
the correlated methods, a large basis set is required to
get the bond lengths right. The rest of the geometries,
including those of the sulfenic acids and alkenes, are
given in the Supporting Information.
CASSCF An a lysis58 of Tr a n sition Sta tes. To probe
the nature of the transition states for diradical character
and ensure that a single-reference description of the
transition states was reasonable, complete active space
self-consistent field (CASSCF) calculations were carried
out on 1TS-4TS and 11TS. For 1TS, 3TS, and 4TS, the
choice of the active space, which consisted of six electrons
in five orbitals, was guided by the illustrated orbital
diagram. Initial orbitals were taken from localized orbit-
als obtained at the HF level and the final orbitals were
well-behaved. Larger active spaces were tested on 1TS.
The starting orbitals included more lone pairs or the SO
σ bond, for instance, but the results either did not provide
(55) Gonzales, C.; Schlegel, H. B. J . Chem. Phys. 1989, 90, 2154-
61.
(56) This justifies the use of the MP2/6-31G(d,p) geometries instead
of the HF/6-31G(d,p) geometries, which are probably on the whole
slightly more accurate for the sulfoxide reactants.
(57) Hargittai, I. In The Chemistry of Sulfones and Sulfoxides; Patai,
S., Rappoport, Z., Stirling, C. J . M., Eds.; J ohn Wiley & Sons Ltd.:
New York, 1988; p 33-53.

8728 J . Org. Chem., Vol. 66, No. 26, 2001
Ta ble 6. Ca lcu la ted ∆Hqelim , ∆Hr xn , a n d ∆Hq
Cubbage et al.
a
a d d n
compd
∆Hq
∆Hrxn
∆Hq
compd
∆Hq
∆Hrxn
∆Hq
elim
addn
addn
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
32.3 (33)
30.4 (30)
30.6 (26.3)b
30.7
30.0
42.7
22.6
20.6
15.7
18.0
20.1
42.6
28.9
20.6
16.6
21.1
28.7
29.9
18.8
21.0
21.0
19.9
16.9
9.7
18
19
20
21
22
23
24
25
26
27
28
29
46.8
46.1
25.8
20.9
21.2
35.5
27.1
23.4
24.6
12.6d
12.4d
17.2d
26.4c
15.6
23.7
17.3
25.3
0.7
8.0
7.1
12.2
4.1
6.4
9.8
14.9
12.7
9.9
0.1
6.2
8.5
9.8
26.7
11.6
12.9
3.2
33.8
28.0
33.4
39.6
33.5
31.2
31.8
32.7 (35)
10.3
33.2
21.3
25.0
24.2
23.2
29.8
35.1
7.8
7.2
29.1 (29)c
27.4 (25-30)c
47.8
20.1
3.5f
16.0
0.1e
9.4
0.5
5.9
4.5
40.3 (42)
42.8
22.0
25.2
25.3
30 en d oe
30 exoe
31 en d oe
31 exoe
4.2
4.3
3.7
2.4
23.6
19.3
a
MP2/6-311+G(3df,2p)//MP2/6-31G(d,p) calculations. All values in kcal/mol and data from experimental analogues are given in
parentheses. Ea, accompanied by surprisingly negative activation entropy. See ref 41 and text. c Experimental values are activation
b
energies Ea. See refs 32 and 33 for experimental values. See text. ∆Hrxn is for free products, not hydrogen-bonded complexes. e Endo
d
and exo refer to endocyclic and exocyclic olefin formation. f ∆Hq
here refers to an intermolecular complex as starting material and
addn
ignores ZPE.
any new information or resulted in final orbitals that
correlated electrons in obviously uninvolved parts of the
molecule. The original [6,5] set was thus taken as the
baseline from which to work on other molecules.
The data for compounds 26-29 deserve special com-
ment, particularly since 27 and 29 appear to have
activation enthalpies that are lower than the total
enthalpy change for the reaction. This is an artifact of
having a transition state whose energy is very close to
the initial product of the reactions, which is a hydrogen-
bonded complex. The values for ∆Hrxn are taken not for
such complexes, but for isolated olefin and sulfenic acids.
IRC calculations were done starting from 26TS-29TS
at MP2/6-31G(d,p), and they led downward in energy to
complexed products. The hydrogen bonded complex for
27 was reoptimized at MP2/6-311+G(3df,2p), along with
For ethyl vinyl sulfoxide 2TS, and methyl vinyl sul-
foxide 11TS, the active space contained eight electrons
in seven orbitals. In addition to the same six electrons
in five orbitals used previously, the π and π* orbitals were
added. For 3TS, an active space of the same size was
used. It included the same basic [6,5] active space, the
highest lying benzene π orbital, and the lowest lying π*.
An important result of these calculations is the natural
orbital occupation numbers (NOONs). In systems that
are reasonably described as closed shell, these are near
0 and 2. For example, the NOONs for ethylene at
CASSCF[4,4]/6-31G(d,p) are 1.982 (C-C σ), 1.920
(C-C π), 0.079 (C-C π*), and 0.018 (C-C σ*). The
NOONs obtained for the current transition states (0.020-
0.092 and 1.911-1.998, see table in the Supporting
Information) are all consistent with a dipolar and not
diradical transition state. These results support the idea
that a single reference method, such as MP2, may be
sufficient to get meaningful results. The transition states
1TS and 11TS were reoptimized at the CASSCF level.
Only insignificant changes were observed in the geom-
etries, energies, and NOONs. Thus, the data in the
Supporting Information are all from CASSCF/6-31G-
(d,p)//MP26-31G(d,p) calculations.
the transition state and starting material. The ∆Hq
elim
increased from 10.3 to 11.4 kcal/mol. The ∆Hrxn, counting
the complex as products, is 7.9 kcal/mol at the MP2/6-
311+G(3df,2p) level of theory. The complex is 5.8 kcal/
mol more stable than the isolated products, methane-
sulfenic acid and formaldehyde, and the O-H hydrogen
bond distance is calculated to be 1.90 Å. Because these
structures were found by IRC methods rather than with
the intention of finding the global minimum, the analo-
gous hydrogen-bonded complex for 26 is a different
“rotamer”, with an O-H distance of 1.94 Å. Similarly, a
hydrogen-bonded structure for the products of thermoly-
sis of 28 was located by IRC methods. Its N-H distance
is 1.85 Å. Geometries are given in the Supporting
Information.
A similar analysis was completed on 29. The IRC (MP2/
6-31G(d,p)) produced a complex associating thioformal-
dehyde and methanesulfenic acid. This was reoptimized
at MP2/6-311+G(3df,2p) and found to be 4.6 kcal/mol
more stable than the free products. The complex, whose
geometry is very similar to the transition state, is only
0.1 kcal/mol below the transition state in the absence of
vibrational ZPE corrections. When zero-point energies are
added in, the transition state disappears as a maximum.
In other words, methanesulfenic acid adds to thioform-
aldehyde to give the thiosulfinate without barrier when
ZPE is included. A lower energy, hydrogen-bonded com-
plex was also found. The overall calculated energetics of
the thermolysis of 29 are summarized in Table 7.
Com p u ta tion a l Resu lts. The energies calculated for
∆Hq and ∆Hrxn, along with the experimental data for the
analogues where available, are shown in Table 6. As
justified in the previous paragraphs, the data are from
MP2/6-311+G(3df,2p)//MP2/6-31G(d,p) calculations. The
∆Hq
values are the activation enthalpies for the
elim
elimination (the “forward” reaction), and the ∆Hq
addn
values are those for the reverse reaction, i.e., the addition
of the sulfenic ester to the olefin to form the sulfoxide.
Finally, a comment should be made about accuracy and
precision. There is not a “random error” associated with

Thermolysis of Alkyl Sulfoxides and Derivatives
J . Org. Chem., Vol. 66, No. 26, 2001 8729
Ta ble 7. En er geticsa of 29 w ith TS, F r ee P r od u cts, th e In itia l Com p lex (29vd w ), a n d th e Hyd r ogen -Bon d ed Com p lex
(29h b)
∆Hq
no ZPE
∆Hrxn
freeb w/ZPE
level
ZPEc
freeb
29vd w
29vd w w/ZPE
29h b
29h b w/ZPE
MP2/6-31G(d,p)
MP2/6-311+G(3df,2p)//MP2/6-31G(d,p)
MP2/6-311+G(3df,2p)
17.2
25.6
25.9
13.6
22.0
23.3
13.3
30.4
31.2
10.3
27.4
28.3
7.4
24.5
25.1
5.8
22.9
23.3
25.8
23.9
a
b
All energies in kcal/mol and are stated relative to the lowest energy conformer of 29. When not specified, ZPE is not included. Free
d
refers to separate calculations for thioformaldehyde and methanesulfenic acid. ZPE energy used from the optimized geometry at the
MP2/6-31G(d,p) level.
Ta ble 8. Com p a r ison of P r im a r y Kin etic Isotop e Effects
the calculations, meaning there are no “standard devia-
w ith Liter a tu r e Mea su r em en ts
tions.” However, the degree of matching to experimental
kH/kD
kH/kD
kH/kD
data and differences observed when using closely related
protocols (e.g., optimizing at the biggest basis set or not)
suggest limits below which the significance of the num-
bers is questionable. For purposes of discussion, we will
look for differences of at least 2 kcal/mol in calculations
to be confident of “real” variation from structure to
structure. Comparison of very closely related systems
may justify discussion of smaller differences, but a degree
of realism is required.
T (K) expt64 3/3-d 1 T (K) expt38 3/3-d 3 T (K) expt32 8/8-d 9
298
4.39
3.07
2.92
2.80
2.68
2.58
2.49
298
5.23
4.14
3.99
3.85
298 (15.6)a 5.81
403 3.17
423 2.94
443 2.77
463 2.63
483 2.49
503 2.38
353 5.15
363 4.97
373 4.77
385
393
400
408
5.4
4.5
3.8
3.4
3.97
3.86
3.78
3.68
a
Extrapolated. See text.
Exp er im en ta l a n d Com p u ted Isotop e Effects.
Kinetic isotope effects (KIEs) were evaluated experimen-
tally for 1e by using 1e-d 4 and 1e-d 6 as substrates. The
kH/kD was determined over the temperature range of
230-280 °C with three runs at each temperature at 10°
intervals. The difference between the KIEs over this
range was less than the scatter over the data, so the
average was taken over the full temperature range. The
kH/kD for 1e vs 1e-d 4 was found to be 2.5 ( 0.3. The
kH/kD for 1e vs 1e-d 6 was found to be 2.8 ( 0.9, where
the error limits are two standard deviations. There is no
good chemical reason to expect a large difference in KIE
between 1e-d 4 and 1e-d 6; only a secondary KIE is
expected. We interpret the experimental data for these
two compounds, as being experimentally indistinguish-
able, given the large error bars.
KIE computations were also carried out to compare to
experimental results reported by other workers (Table
8). Calculated isotope effects for 3-d 1 and 3-d 3 are shown
at room temperature and the experimentally relevant
temperatures. The calculations for 3-d 1 nicely match
Kwart’s data for 3K, collected in diglyme.64 That conjuga-
tion in the transition state is important (vide infra) is
demonstrated by comparing the KIEs calculated for 3-d 3
to the compound 3Y, studied in dioxane by Yoshimura.38
The poor match clearly implies that data calculated for
3 are much less relevant to 3Y than to systems without
phenyls or other conjugating substituents.
KIE computations were done on model compounds 1,
1-d 2, and 1-d 4 to attempt to model the KIE and further
support the calculated transition-state model. The pro-
gram ISOEFF98,59,60 which uses vibrational frequencies
from the substrates and respective transition states to
solve for the KIE using the Bigeleisen equations,61-63 was
used in conjunction with the vibrational output from
GAMESS. At 298 K, the calculated KIEs are 4.6 and 5.0
for 1-d 2 and 1-d 4, respectively. This submaximal value
is reflective of the somewhat bent geometry of the proton
transfer. Over the temperature range of the experiments
(503-553 K) both 1-d 2 and 1-d 4 have calculated KIEs
that average to 2.5, in excellent agreement with the
experiments.
The experimental isotope effects32 for 8 are somewhat
unusual. Kwart interpreted32,65 the AH/AD of 0.07 and ∆Ea
of 3.2 kcal/mol as indicative of a mechanism that included
a hydrogen atom tunneling component attributable to
sterically induced narrowing of the barrier. Extrapolated
to 298 K, the KIE for Kwart’s compound is 15.6. The
computational method used here explicitly excludes
tunneling effects and cannot possibly duplicate these
KIEs, regardless of the accuracy of the characteristics of
the calculated transition states or whether tunneling is
actually involved. It should be noted that the experimen-
tal and computed activation enthalpies for elimination
by 8 are in excellent agreement (Table 6), though this
does not eliminate the possibility of tunneling near the
top of the barrier. For the sake of completeness, the
computational data are included in Table 8.
(58) Schmidt, M.; Gordon, M. S. Annu. Rev. Phys. Chem. 1998, 49,
233-66.
(59) Anisimov, V.; Paneth, P. J . Mathem. Chem. 1999, 26, 75-86.
(60) Paneth, P. Computers Chem. 1995, 19, 231-40.
(61) Bigeleisen, J .; Mayer, M. G. J . Chem. Phys. 1947, 15, 261-7.
(62) Wolfsburg, M. Acc. Chem. Res. 1972, 6, 225-33.
(63) Melander, L. Isotope Effects on Reaction Rates; Ronald Press:
New York, 1960.
(64) Kwart, H.; George, T. J .; Louw, R.; Ultee, W. J . Am. Chem. Soc.
1978, 100, 3527-8.
(65) Kwart, H. Acc. Chem. Res. 1982, 15, 401-8.

8730 J . Org. Chem., Vol. 66, No. 26, 2001
Cubbage et al.
Ta ble 9. Bon d Or d er In d ex Va lu esa for Rep r esen ta tive
Su lfin yl Der iva tives, Tr a n sition Sta tes, a n d P r od u cts
Instead, we report the MP2/6-31G(d,p) bond order
indexes (BOIs)66-68 for the transition states and certain
representative sulfoxides in Table 9. Inspection of Table
9 will reveal that the BOIs for ordinary bonds are
typically slightly less than 1.0. The C-H and C-C bonds
in 1 are both over 0.95, while the C-S bond is 0.86, which
is quite representative of most of the reactants. The S-O
bond order of 1.36 reflects the additional “bonding” that
results from the electrostatic attraction of the ylide but
also points out that there is not a classic second bond
between the two atoms. For purposes of further discus-
sion, we will not consider differences in BOI significant
unless they approach 0.05.
compd
S-O
O-H
H-Y
X-Y
X-S
1
1TS
CH3SOH
2TS
3TS
4TS
5TS
6
6TS
F SOH
7TS
H2NSOH
8TS
9
9TS
1.36
1.06
0.82
1.07
1.07
1.11
1.09
1.56
1.18
0.92
1.09
0.81
1.04
1.24
1.03
1.37
1.04
1.04
1.36
1.05
0.96
0.43
0.97
1.33
0.86
0.46
0.41
0.86
0.38
0.38
0.34
0.36
0.45
0.45
0.48
0.47
0.96
0.36
1.31
1.32
1.29
1.30
0.95
1.44
0.47
0.45
0.44
0.48
0.88
0.36
0.45
0.84
0.43
0.86
0.39
Discu ssion
0.40
1.34
0.45
Wh ith er P seu d op er icyclic. Overwhelming experi-
mental and computational evidence shows that the
elimination reaction of sulfoxides is concerted in the vast
majority of cases. It has sometimes been characterized
as pericyclic. However, there is little if any experimental
evidence of a classic double bond between sulfur and
oxygen.69 A coordinate bond, which we have represented
as an ylide, is more in line with experiment and theory.
The most recent Atoms-in-Molecules analysis of the
sulfoxide bond16 sees it as essentially a single bond that
is strengthened and shortened by electrostatic attraction.
Localized molecular orbital views of the oxygen lone pairs
show some distortion back toward the sulfur, consistent
with this picture. If the essence of the enhanced single-
bond view is accepted, then there is no longer reason to
consider this reaction pericyclic, and the analogy to the
Cope elimination holds.
The question of whether the reaction can be called
pseudopericyclic is more interesting. Lemal’s prototypical
case26 involved the automerization of a bicyclic sulfoxide
and the interchange of the S lone pair with a C-S
bonding orbital. Birney and co-workers have recently
shown that many pseudopericyclic reactions occur through
approximately planar transition states with lower barrier
than a corresponding pericyclic, nonplanar transition
state would have.27 In this planarity, and in the exchange
of lone pair and bonding orbitals, Birney’s pseudoperi-
cyclic reactions have much in common with the current
elimination reaction. In a sense, however, the important
feature of a pseudopericyclic reaction it that it is not
pericyclic, but merely is a concerted reaction in which a
cyclic array of action occurs. In that sense, pseudoperi-
cyclic reactions are merely a subset of the many concerted
unimolecular reactions that exist. The syn-elimination
of sulfoxides is a concerted unimolecular reaction, but if
we take seriously the idea of a coordinate, ylide-type
sulfoxide bond, prudence suggests we not label the
reaction pseudopericyclic because the oxygen is assigned
the role of base, the sulfur acts as leaving group, and
the bonding change between S and O is only marginal.
Ad equ a cy of Mod elin g a n d th e Na tu r e of th e
Tr a n sition Sta te. Gas-phase experimental data were
collected here in order that computations could be done
without reference to solvent modeling or other complica-
0.45
0.96
0.47
0.95
0.38
0.39
0.94
0.38
1.29
0.98
1.28
1.93
2.38
1.37
0.95
1.17
1.88
0.95
1.24
1.32
1.30
1.34
0.93
1.44
1.31
1.30
1.27
1.39
1.31
1.31
1.30
0.81
1.41
0.92
1.20
0.90
1.29
0.94
1.44
1.27
1.35
1.29
1.34
0.47
0.86
0.45
0.88
0.50
0.43
0.86
0.61
0.37
11
11TS
12TS
13
13TS
acrolein
14
14TS
15TS
16TS
17TS
18
18TS
19TS
20TS
21TS
22TS
23TS
24TS
25TS
26
26TS
27
27TS
28
28TS
29
29TS
30TS-en d o
30TS-exo
31TS-en d o
31TS-exo
0.47
0.43
0.45
1.35
1.11
1.06
1.08
0.96
1.39
0.97
1.06
1.05
1.08
1.03
1.05
1.04
1.05
1.50
1.08
1.26
1.06
1.41
1.09
1.50
1.03
1.07
1.07
1.07
1.08
0.97
0.48
0.42
0.46
0.23
0.97
0.28
0.43
0.45
0.48
0.35
0.41
0.44
0.44
0.96
0.34
0.79
0.35
0.96
0.44
0.94
0.33
0.46
0.43
0.43
0.44
0.86
0.52
0.43
0.43
0.52
0.88
0.38
0.48
0.48
0.50
0.48
0.48
0.44
0.47
0.77
0.31
0.87
0.70
0.83
0.43
0.83
0.41
0.49
0.41
0.48
0.42
0.36
0.40
0.37
0.62
0.56
0.40
0.39
0.35
0.49
0.42
0.39
0.39
0.48
0.08
0.43
0.40
0.49
0.37
0.39
0.39
0.38
a
The minimum threshold to list a BOI is 0.05.
Tr a n sition Sta tes. With the exception of compound
10, all of the transition states presented here are five-
membered ring cycles and essentially planar. (Compound
10 was specifically included as a vinylogue of the main
reaction and, therefore, has a seven-membered ring cyclic
transition state.) This similarity suggests that compari-
son of some physical parameter might be made among
all the transition states in order to examine which are
“earlier” or “later” and the general synchronicity of the
bond making and breaking. For example, Figure 2
suggests that, for 1, the C-S bond is broken early in the
reaction sequence, whereas the C-H bond breaking lags
behind. Comparison of 1TS and 11TS in Figure 3
suggests that the proton transfer is closer to complete
for 11TS, whereas the C-S bond breaking is similar to
the previous case. However, comparison of geometries
such as this can be awkward and tenuous.
(66) Giambiagi, M.; Giambiagi, M.; Grempel, D. R.; Heymann, C.
D. J . Chim. Phys. Phys.-Chim. Biol. 1975, 72, 15-22.
(67) Mayer, I. Chem. Phys. Lett. 1983, 97, 270-4.
(68) Mayer, I. Chem. Phys. Lett. 1985, 117, 396.
(69) Shorter, J . In The Chemistry of Sulfoxides and Sulfones; Patai,
S., Rappoport, Z., Stirling, C. J . M., Eds.; Wiley: New York, 1988; pp
483-541.

Thermolysis of Alkyl Sulfoxides and Derivatives
J . Org. Chem., Vol. 66, No. 26, 2001 8731
tions. As it turns out, the current experimental data are
in line with previously reported results, both in solution
and gas phase. The calculated activation enthalpies are
all in good agreement where experimental values are
available, and in the cases where only rates at single
temperatures are known, the calculated activation en-
thalpies appear reasonable.
As established in Tables 3 and 4, it is a demanding
problem to compare the energies of sulfoxides and their
isomers, which, broadly speaking, includes the transition
states. From Table 3, it appears that the basis set limit
has been met for the transition state, relative to the
sulfoxides, and the match of the MP2 calculations to the
experimental data support that conclusion.
The use of MP2 calculations is clearly a compromise
in methodology. CCSD(T) calculations were impractical
given the large basis sets needed and the size of the
molecules. However, the behavior of the CASSCF calcu-
lations and the NOONs obtained from them were at least
consistent with MP2 being a reasonable method. Again,
the performance of the MP2 calculations, including the
accuracy of the calculated KIE for 1, argues that the
essence of the transition state is captured in the calcula-
tions.
The performance of the calculations is less clear for
the calculated ∆Hrxn values (i.e., the sulfoxide compared
to the sulfenic acid and olefin) because the convergence
of the data with respect to basis sets is not demonstrated.
However, since we are studying a series of closely related
reactions, it is likely that any error would be systematic
and result in an offset in the data that would be similar
for all the compounds.
in favor of computational chemistry; however, the per-
formance of G2 calculations for molecules of this size is
excellent. Among the complete G2 test set, the single
worst case is SO2, whose calculated heat of formation is
5 kcal/mol different than the accepted experimental
value. Methanesulfenic acid is the same size as SO2 and
has none of the complications of dative/hypervalent
bonding. It is hard to believe that the G2 calculations
would be so much worse for CH3SOH than for any other
molecule its size. When the G2 value for methanesulfenic
acid is used as a surrogate for the experimental value,
the new “experimental” estimate for ∆Hrxn (298 K) for 1
is 21 kcal/mol, much more in line with the 22.6 kcal/mol
(O K) calculated value. It may reasonably be argued that
this can be expected given the consistency of basis set in
our calculations and the G2 calculation, but we suggest
that the reported heat of formation for CH3SOH be
treated with caution.
Gen er a l In ter p r etive Su m m a r y. The current data
are all consistent with a concerted mechanism for the
syn-elimination of sulfoxides and, thus, the addition of
sulfenic esters to olefins. The concertedness of the
elimination, however, does not imply the synchronicity
of events. It is self-evident from Figure 2 that in the
elimination reaction of 1 the breakage of the C-S bond
begins earlier than does the breakage of the C-H bond.
At the transition state itself, however, the BOI data in
Table 9 show that the C-H and C-S bonds are broken
to a similar extent. Given further that a BOI of 1.09
represents the halfway point for the transition of the S-O
bond from sulfoxide to sulfenic ester, it can be concluded
that the transition state for the Ei elimination of 1 is
neither particularly early nor late and that it is roughly
symmetrical with respect to bond making and breaking.
Only the C-C BOI differs significantly from the middle
of the starting and end points, suggesting that the π-bond
formation is lagging.
We were hampered in our efforts to critically evaluate
∆Hrxn because there are so few ∆Hf° data available for
sulfenic acids. Neither are extensive data available for
sulfoxides, but it is reasonable to use the Benson addi-
tivity method for simple alkyl derivatives to get estimates
within a couple of kcal/mol.70
Turecek reported a value of -45.4 kcal/mol for the
∆Hf°(298 K) of CH3SOH.71 Using this, the well-estab-
lished experimental value for ethylene, and a Benson-
If the elimination reaction is naturally pictured as
using the sulfinyl oxygen acting as an internal base while
the sulfinyl sulfur acts as leaving group, then the reverse
addition can be seen as having the sulfur act as nucleo-
phile while the sulfenic acid proton is removed by the
π-bond acting as a Lewis Base. This is consistent with
experimental results in which activated olefins or acet-
ylenes have been used as traps for sulfenic acids. The
regioselectivity of that addition reaction always favors
the sulfoxide product that is formed when sulfur attacks
the more electrophilic carbon of the olefin, i.e., with
Markovnikov selectivity.50
type estimate for 1, one obtains an estimate of ∆Hrxn
)
9 kcal/mol. Perhaps coincidentally, this is approximately
the value computed at all levels of theory using the
6-31G(d,p) basis set (Table 4). However, when the better
MP2/6-311+(3df,2p) calculations were done, suddenly the
MP2 computed value of 22.6 kcal/mol is in terrible
disagreement with our “experimental” estimate of 9 kcal/
mol.
There are several trends to be seen within the current
data. For some compounds (e.g., 6, 18), the ∆Hrxn is much
larger than the rest because of particularly unstable
products. In those cases, the transition states are natu-
rally product-like. However, among the more closely
related sulfoxides, the variation in ∆Hqelim values is small.
This observation is in line with previous more extensive
experimental reports.30,37,38 Moreover, at first glance, the
This led us to reexamine the heat of formation of CH3-
SOH using a G2 calculation.72-74 Using either the atomi-
zation technique (∆H°f ) -33.5 kcal/mol) or isodesmic
reactions (-33.6 kcal/mol), the G2 heat of formation at
298 K was 12 kcal/mol different than Turecek’s experi-
mental value. We are loathe to reject experimental values
variation in ∆Hq
seems to track with ∆Hrxn. On closer
(70) Stein, S. E.; Lias, S. G.; Liebman, J . F.; Levin, R. D.; Kafafi, S.
A.; NIST Structures and Properties: NIST Standard Reference Data-
base 25, 2.0 ed.; U.S. Department of Commerce, NIST: Gaithersburg,
MD, 1994.
(71) Turecek, F.; Brabec, L.; Vondrak, T.; Hanus, V.; Hajicek, J .;
Havlas, Z. Collect. Czech. Chem. Commun. 1988, 53, 2140-58.
(72) Curtiss, L. A.; Carpenter, J . E.; Raghavachari, K.; Pople, J . A.
J . Chem. Phys. 1992, 96, 9030-4.
(73) Curtiss, L. A.; Raghavachari, K.; Redfern, P. C.; Pople, J . A. J .
Chem. Phys. 1997, 106, 1063-79.
(74) Curtiss, L. A.; Raghavachari, K.; Trucks, G. W.; Pople, J . A. J .
Chem. Phys. 1991, 94, 7221-30.
elim
inspection, though, there are interesting patterns among
the ∆Hq
data that may be more straightforward to
addn
interpret than the variations in ∆Hqelim. For example, the
compounds most closely related to the standard bench-
mark 1 have ∆Hq
of approximately 10 kcal/mol. Even
addn
compounds 9, 11, and 12 are in this ballpark. This value
appears to be characteristic of systems in which neither
the olefin nor sulfenic acid are activated. In other words,

8732 J . Org. Chem., Vol. 66, No. 26, 2001
Cubbage et al.
variations in the ∆Hq within this group are mainly due
to little more than changes in overall endothermicity.
(i.e., derivatives of 3) is still prohibitively demanding of
computer resources.
elim
Interpretation of the ∆Hq
data can be taken further.
The carbonyl-conjugated systems find in themselves a
new group with small ∆Hqaddn (3-5 kcal/mol), presumably
due to conjugation in the transition state. Polarity
matters, as with 13 vs 14, but the effect is small, relative
to the mere existence of a more delocalized transition
state. Nonetheless, as Markovnikov-type selectivity is
always observed experimentally, it is reassuring to see
addn
When the sulfenic acid sulfur is made less nucleophilic,
as in the product from compounds 3 and 4, ∆Hq
is
addn
higher. When the opposite is true, e.g., 7, there is a
decrease in ∆Hqaddn. (We presume that the R-effect of the
nitrogen in 7 increases the sulfur nucleophilicity. The
oxygens of sulfinamides are also known to be less basic
than sulfoxides.75-77) With compound 17, the olefin is
activated by the ammonium substituent, making it more
electrophilic and lowering ∆Hqaddn. In the forward direc-
tion, the ammonium substituent can be seen as increas-
ing the acidity of the abstracted proton and is the only
example in which additional conjugation in the transition
state is not introduced along with the change in acidity.
at least small differences in ∆Hq
that favor the
addn
Markovnikov product. Carbonyl compounds themselves
are obviously electrophilic and very strongly polarized.
In the absence of extra conjugative effects, the polariza-
tion becomes very important, as, for example, the ∆Hq
addn
for 26 is larger than average and that of 27 is very small.
Returning to the elimination direction, this can also be
seen as an acidity effect, where the OH of 27 is quite
acidic compared to the analogous protons in benchmark
compounds.
Electron-withdrawing (and therefore electron-releas-
ing) substituents might in principle exert effects in both
favorable and unfavorable directions with respect to the
elimination reaction. The oxygen atom acts as an internal
base, and thus, electron-withdrawing substituents on the
nonreactive side of the sulfoxide might be expected to
raise the activation barrier by decreasing39 basicity.
However, there is a countervailing structural issue. The
S-O charge separation is greater in the sulfoxide than
in the sulfenic acid product. As a qualitative benchmark,
the MP2/6-31G(d,p) Mulliken charges on sulfur for 1,
1TS, and CH3SOH are 0.82, 0.57, and 0.38, respectively.
Therefore, electron-withdrawing substituents might also
preferentially destabilize the sulfoxide over the transition
state and sulfenic acid, thus lowering the barrier.
Gr ou p s 1 a n d 2: Su lfoxid es w ith Su bstitu en ts on
th e Non r ea ctive Sid e. Given the prevailing wisdom in
the synthetic use of the sulfoxide elimination that a
phenyl substituent causes the reaction to go at lower
temperature than a methyl, we chose phenyl ethyl
sulfoxide (3) as a model experimental system to compare
with 1. The difficulty in getting data (and the relative
size of the molecule for computational study) made 2 a
logical compromise. The observed differences in activa-
tion barrier between 1 and 2 are relatively small, but in
the end served as an important test for the computational
method.
We believe the Hammett plot results of Emerson and
Yoshimura are interpretable within this context, using
4 as a model. The small Hammett F values are consistent
with there being the two effects in opposite directions,
especially given the relatively large F value of opposite
sign known for acidity effects on aryl methyl sulfoxides.39
Relative to 1, the barrier for elimination is lower and the
barrier for addition for 4 is higher. This is a consequence
of the electron-withdrawing CF3 substituent having a
destabilization influence in the order sulfoxide > transi-
tion state > sulfenic acid, consistent with the charge
argument. It thus must be the case that the decrease in
basicity of the sulfoxide is more than made up for in the
transition state by the presence of the electron-with-
drawing group next to the partially positive sulfur. In
more “chemical” language, we can speak of this being the
drop in nucleophilicity of the sulfenic acid sulfur or the
increased nucleofugacity of the sulfoxidic sulfur, with a
smaller effect on the basicity of O. Both Emerson and
Yoshimura showed a small decrease in activation en-
thalpy for nitrophenyl sulfoxides (X ) NO2 in Scheme
1), relative to the parent phenyl sulfoxide.
Experimental data are also available for phenyl ethyl
sulfoxide 3.41 In dioxane solution, an activation energy
(Ea) of 26.3 kcal/mol over an unspecified apparently
20 °C range was obtained, along with a very negative
activation entropy of -18 ( 4 cal/mol‚K that is the most
negative reported in the study. It is at least coincidental
that such a very negative activation entropy accompanies
an activation energy that we apparently overestimate
computationally, especially given our experimental data
for 2e. The reasons for this discrepancy are not clear.
The molecules in group 2 were chosen in an attempt
to find a system where there might be a greater variation
in ∆Hq
and to test the ideas about countervailing
elim
effects of substituents on basicity and nucleofugacity of
the sulfinyl group. As discussed above, the activation
barrier for 4 is consistent with destabilization of the
sulfur position being the predominant effect in the
sulfoxide. This may be reflected in the comparatively
incomplete proton transfer in the transition state as
illustrated in the BOI values. Qualitative analysis of the
BOIs in the addition direction is also ambiguous in that
the proton of CF3SOH is certainly more acidic than that
of CH3SOH, but the sulfur is less nucleophilic.
The opposite energetic effect would be expected for
electron-rich substituents, that is, a differential stabiliza-
tion in the order of sulfoxide > transition state > sulfenic
acid. Again, the experimental Hammett results are in line
with this interpretation. Unfortunately, carrying out a
series of calculations at the MP2/6-311+G(3df,2p)//MP2/
6-31G(d,p) level for substituted phenyl ethyl sulfoxides
With 7, analysis of the addition reaction is more
straightforward. The electronegativity of the nitrogen,
relative to carbon, should make H2NSOH more acidic
than methanesulfenic acid. However, the R-effect of the
lone pair on N is also expected to contribute to the
nucleophilicity of sulfur. Thus, as observed, a lower-than-
benchmark ∆Hq
is expected. This is a prime example
can be difficult to interpret in isolation
addn
of how the ∆Hq
(75) Henrickson, C. H.; Nykerk, K. M.; Eyman, D. P. Inorg. Chem.
1968, 7, 1028-9.
elim
from other data. It is higher than the benchmark value
at least largely because of the high ∆Hrxn. The appearance
of 7TS is unremarkable, save for being behind in making
(76) Engberts, J . B. F. N.; Hovius, K.; Zuidema, G. Rec. Trav. Chim.
Pays-Bas 1971, 90, 633-40.
(77) Ruostesuo, P. Finn. Chem. Lett. 1979, 202-5.

Thermolysis of Alkyl Sulfoxides and Derivatives
J . Org. Chem., Vol. 66, No. 26, 2001 8733
the OH bond (in the elimination direction), consistent
with the diminished basicity75-77 of the oxygen in sulfin-
amides.
In compound 6, there is a case in which the elimination
is so endothermic (43 kcal/mol, compared to 23 for 1) as
to cause the transition state to be very late and product-
like.78 The lateness of 6TS is reflected in the BOIs, but
they are also more asymmetric with respect to bond
making and breaking, with the fluorine reducing the
basicity of oxygen and hindering proton transfer.
Gr ou p 3: Ster ica lly Hin d er ed Su lfoxid es. Among
these two sulfoxides, the greater effect is expected for di-
tert-butyl sulfoxide 9, in which greater steric crowding
of the starting material is relieved by the elimination
reaction.33 As a gauge of the steric strain, the calculated
C-S-C angle for 9 is 15° greater than the experimental
value for DMSO.57 Mulliken charge analysis points to a
slightly more negative O than usual (-0.84 vs -0.80 for
1), and this is accompanied by a notably smaller S-O
BOI than usual (1.24 vs 1.36 for 1).
F igu r e 4. Top and side view of 10TS.
and calculation in Table 6, the striking results are the
very similar values of ∆Hq
that 11 and 12 share with
addn
1. In other words, the addition is not particularly
activated by the sp center, compared to the usual sp2
center.
Gr ou p 5: Ca r bon yl-Con ju ga ted Olefin P r od u cts.
This group of compounds was used to examine the effect
of a carbonyl-conjugated product, as found in the most
common synthetic applications of this chemistry. Com-
pounds 13-16 are marked by having unremarkable
∆Hrxn values that are a little lower than the parent
compound 1 because of the conjugated carbonyl in the
Shelton presumed that C-S bond cleavage would come
early along the reaction pathway to relieve strain. Kwart
suggested that the narrower barrier inherent in the steric
crowding might lead to increased tunneling.32,65 The
simplest explanation consistent with the current data is
that steric strain destabilizes the starting material
without causing a major shift in the location of the
product. They have significantly lowered ∆Hq
values,
elim
all in the range of 22-25 kcal/mol and correspondingly
low ∆Hq values. The “vinyl effect” (comparing ∆Hq
addn
elim
for 1 and 2) of lowering activation enthalpy by about 2
kcal/mol is reproduced between 15 and 16. As is reason-
able, there is little difference between the aldehyde and
ketone substituents used in 14 and 15. These lower
activation energies are at least qualitatively in line with
experiment in that the polyfunctional sulfoxide shown
below was shown to racemize in boiling chloroform by
way of the reversible elimination reaction, whereas very
similar sulfoxides without the carbonyl were stable.81
transition state. The ∆Hq
values for 8 and 9 are very
addn
similar to that for 1, indicating that the majority of the
strain in 8 and 9 is relieved in the transition state. The
BOIs of the transition state for 9, interestingly, are not
biased toward greater than usual breakage of the C-S
bond, but in fact is early with respect to O-H bond
formation. Though we cannot model a transition state
in which tunneling is significant, the greater O-H bond
formation in the classical transition state is at least
consistent with Kwart’s proposal that tunneling is sig-
nificant in these cases.65
Gr ou p 4: Su lfoxid es w ith Dou ble Bon d s. The
activation enthalpy for the vinylogous analogue 10 is
higher than for the other compounds. The ∆Hq
value
addn
is also higher than usual, meaning that this is not due
to a simple endothermicity problem. (This substrate could
not be studied experimentally because of the competing
reversible rearrangement of allyl sulfoxides to allyl
sulfenic esters.79) We attribute this to the strain inherent
in the seven-membered ring transition state.80 In Figure
4, it can be seen that although the general alignment of
the five key atoms is about the same as usual, albeit with
the orthogonal diene extension in the middle, there is a
distinct distortion of those five atoms from planarity.
The other members of this group, 11 and 12, test what
happens when the abstracted proton is attached to an
Compound 14 has the regiochemistry produced in the
usual synthetic sequence when the sulfoxide is made
from an enolate, but 13 is an isomer in which the sulfinyl
oxygen is set up to remove the more acidic proton.
Compound 13 is also the isomer that would be observed
as the major product from addition of methanesulfenic
acid to acrolein. The difference in ∆Hrxn between the two
reflects a 2.2 kcal/mol greater stability of 14 and is
reflected in a 1 kcal/mol lower barrier to form 14. The
clear message from the energetics is that the major part
of the differences between these compounds and 1 is due
to the carbonyl conjugation in the transition state, with
the polarity/acidity playing a smaller part. These results
are in line with the experimental observations of Crich,
who used esters instead of aldehydes.40
Tsukurimichi came to the same conclusion that car-
bonyl conjugation in the transition state accelerated the
rate of elimination, and the reported activation energies
(Ea) for phenyl 1-carbomethoxy-1-methylethyl phenyl and
2-carbomethoxyethyl phenyl sulfoxides are both within
experimental error of 26 kcal/mol.41 In this same paper
is reported the surprisingly low barrier for 3 (vide supra),
sp2 carbon. The less negative experimental ∆Sq
(Table
elim
2) reflects the fewer degrees of freedom in the starting
material because of the restricted rotation of the C-C
bond. Aside from the excellent match between experiment
(78) It should also be noted that computational modeling of the O-F
bond is not straightforward and requires more sophisticated treatments
than given here (e.g., CCSD(T)). We have not critically evaluated the
F-S bonds calculated here, so some caution should be used in
conjunction with the absolute energies given in the tables. Lee, T. J .;
Rice, J . A.; Dateo, C. E. Mol. Phys. 1996, 89, 1359-1372.
(79) Tang, R.; Mislow, K. J . Am. Chem. Soc. 1970, 92, 2100-4.
(80) Beginning with S-O and going around the cycle of the transi-
tion state, the BOIs are 0.98 (S-O), 0.48 (O-H), 0.34 (H-C), 1.30 (C-
C), 1.48 (C-C), 1.16 (C-C), 0.67 (C-S).
(81) Stoodley, R. J .; Wilkins, R. B. J . Chem. Soc., Perkin Trans. 1
1974, 1572-9.

8734 J . Org. Chem., Vol. 66, No. 26, 2001
Cubbage et al.
which if actually somewhat higher, would seem to be
more consistent with their conclusion.
35 °C for 32.84 If ∆Hq
is taken at the calculated value
elim
of 20 kcal/mol, the resulting ∆Sqelim is -5 cal/mol‚K. Greer
and Foote found an experimental activation energy for
the closely related 2-methyl-2-butene episulfoxide of 18
kcal/mol.85
The complementary nature of structures 13 and 14 is
reflected well in the corresponding transition-state struc-
tures. The structure of 13TS, in which the partial
conjugation is to the â-carbon whose proton is being
abstracted, is considerably later than benchmark for
O-H formation and C-H breaking, while it is rather
early with respect to C-S breaking, where there is no
conjugative advantage. Relative to 1, 14TS is early in
every respect as measured by the BOIs. However, relative
to 13TS, it is late with respect to C-S dissociation, i.e.,
relatively speaking, the C-S bond, whose breakage is
assisted by the conjugative effects of the carbonyl occurs
before the proton abstraction. The C-S distance (see the
Supporting Information) in 14TS is fully 0.25 Å longer
than that of 13TS. These effects are in line with the
suggestion of Yoshimura et al., based on rate data at fixed
temperatures, that a â-electron-withdrawing group will
encourage what they termed a “carbanion-like” mecha-
nism in which the proton transfer proceeds in advance
of C-S bond breaking.82
Gr ou p 7: Heter oa tom ic Su lfin yl Der iva tives. The
calculated activation enthalpy for 26 of 32.7 kcal/mol
compares well with the experimental value of 34.6 ( 0.6
for 26e. To the best of our knowledge, this is the first
report of formation of an aldehyde by elimination from a
sulfinic ester. The calculated ∆Hrxn for this reaction is 5
kcal/mol smaller than that for 1. Given the very similar
∆Hrxn calculated for the isomeric sulfoxide 27, the low
∆Hrxn value of 12.4 kcal/mol derives from the favorable
formation of a carbonyl π system, rather than instability
of the sulfinic ester.
Hypothetical compound 17 demonstrates that the
olefin polarity/proton acidity effect can be considerable
The large difference in ∆Hq
between 26 and 27 is
elim
due to the energy difference between the transition states
themselves. In fact, 27TS is over 20 kcal/mol more stable
than the isomeric 26TS. This is easily understood by
examining the addition reaction. The much larger barrier
to addition in the sense that leads to 26 reflects the
reversal of polarity of the carbonyl. The nucleophilic
sulfur attacks the electrophilic carbon only on the path
to 27 and the effect is quite substantial.
in the absence of extended conjugation. Its ∆Hq
is
addn
among the very lowest in the whole group. The BOIs
show the expected asymmetry if it is taken that the
carbon acidity is responsible for the relative ease of
elimination. The S-O, O-H, and C-H BOIs are ex-
tremely product-like in 17TS, whereas the C-C BOI is
unremarkable and the C-S BOI is, if anything, some-
what sulfoxide-like.
Gr ou p 6: Rin g Str a in a n d En d ocyclic vs Exocy-
clic Olefin for m a tion . Consistent with the experimen-
tal results of Kice35 and Yoshimura,36 we find a lower
enthalpic barrier for the formation of cyclopentene (20)
than for cyclohexene (21). The awkwardness of 21TS,
because of the need of the ring to flatten out, is confirmed
by the relatively high ∆Hq
for 21 compared to 25. The
This analysis suggests that 27TS ought to have a
considerable C-S BOI, especially compared to the cor-
responding S-O BOI in 26. This is borne out by the data.
The C-S BOI of 27TS is 0.70, the highest among all the
transition states. The analogous C-S BOI for 26TS is
0.31, which is the lowest.
The contrast between 26 and 27 can also be seen in
the elimination reaction, using the idea of the acidity of
the abstracted proton as being determining. The pKa of
the hydroxylic proton of 27 in water is obviously lower
than the corresponding proton in the slower-reacting 26,
but gas-phase acidities are more relevant. A reasonable
estimate of the gas-phase acidity of the hydroxylic proton
in 27 of 370 kcal/mol (∆H°) can be made from data for a
series of alcohols.86 Data for sulfinic esters are not
available, but the gas-phase acidity of the methyl protons
in methyl formate is 391 kcal/mol.87 The value for 26
should be similar. Clennan has proposed that R-hydroxy-
sulfoxides are transient intermediates formed in the
reaction between 1O2 and alkyl sulfides.88,89 The very
addn
geometry can be obtained from the Supporting Informa-
tion. However, the current models are not adequate to
predict the experimental product mixture of 24:1 endo/
exo methylcyclohexene obtained at 100 °C from the
elimination of 1-methylcyclohexyl phenyl sulfoxide.36
Yoshimura et al. also investigated the exo/endo selectivity
for 4-tert-butyl-1-methylcyclohexyl phenyl sulfoxides,83 in
which the axial/equatorial position of the sulfoxide is
locked, but molecules of this size are beyond the scope of
the present investigation.
The formation of highly strained cyclopropene (18) or
methylenecyclopropane (22) raise ∆Hrxn, and a cor-
respondingly late transition state is observed. The cy-
clobutyl compounds show apparently ordinary ∆Hq
elim
and ∆Hrxn, but closer inspection shows that some relief
of ring strain is observed in the ∆Hq
values.
addn
An additional alicyclic ring opening was calculated, for
32, in which ring strain is relieved by the elimination
reaction. The ∆Hrxn is only 0.5 kcal/mol and the ∆Hq was
calculated to be 20.0 kcal/mol using the usual MP2/6-
311+G(3df,2p)//MP2/6-31G(d,p) methodology. This value
is quite reasonable, given Baldwin’s half-life of 3 min at
(84) Baldwin, J . E.; Hofle, G.; Choi, S. C. J . Am. Chem. Soc. 1971,
93, 2810-2.
(85) Greer, A. Personal communication, 2001.
(86) Carroll, F. A. Structure and Mechanism in Organic Chemistry;
Brooks/Cole: Pacific Grove, 1998.
(82) Yoshimura, T.; Yoshizawa, M.; Tsukurimichi, E. Bull. Chem.
Soc. J pn. 1987, 60, 2491-6.
(83) Yoshimura, T.; Sekioka, T.; Shimasaki, C.; Hasegawa, K.;
Tsukurimichi, E. Nippon Kagaku Kaishi 1993, 370-4.
(87) DePuy, C. H.; Grabowski, J . J .; Bierbaum, V. M.; Ingemann,
S.; Nibbering, N. M. M. J . Am. Chem. Soc. 1985, 107, 1093.
(88) Clennan, E. L.; Toutchkine, A. J . Am. Chem. Soc. 2000, 122,
1834-5.

Thermolysis of Alkyl Sulfoxides and Derivatives
J . Org. Chem., Vol. 66, No. 26, 2001 8735
modest calculated barrier to elimination is consistent
with the transient nature hypothesized there.
transition states. In the two endo transition states, the
S-O bond and carbonyl are pointed in essentially op-
posite directions, thus having approximately opposing
dipoles. In the two exo transition states, the lactone
dipole and sulfinyl dipole are disposed at roughly right
angles. In both the endo and exo cases, the lower energy
transition states (i.e., 30TS-exo and 31TS-en d o) have
the sulfinyl methyl over the carbonyl carbon, instead of
over a methylene (31TS-exo) or the exocyclic methyl
(30TS-en d o), as can be seen in the Supporting Informa-
tion. Thus, it would appear that the TS energy depends
not only on the dipoles, but also on fairly subtle steric
interactions.
The elimination reaction energetics of 28 follow the
pattern established by 26, though there are obviously
quantitative differences. Thiosulfinate 29 produces thio-
formaldehyde with a barrier of 21.3 kcal/mol. The weak
S-S homolytic bond strength of 28 (ca. 45 kcal/mol90)
certainly contributes to the low barrier in the forward
direction, and the weak π-bond in the product contributes
to the low barrier in the reverse direction. Most notable
of this reaction is the flatness of the reaction potential
in the vicinity of the transition state when the IRC is
computed at MP2/6-311+G(3df,2p). (See the Supporting
Information.) The transition state 28TS is closer in BOIs
to the benchmark 1TS than to 27TS, but this should be
treated with caution, given the flatness of the potential.
Although free products are higher in energy than the
transition state, we found an initial van der Waals type
complex and a hydrogen-bonded complex as described
previously.
Su m m a r y a n d Con clu sion s
The stirred-flow experiments have allowed access to
activation parameters for a variety of sulfoxides and
sulfinyl derivatives. The activation enthalpies are well
modeled by an MP2 calculation with a sufficiently large
basis set. Tests using CCSD(T) and CASSCF methods
did not indicate that there was need for a multireference
correlated method for the computations. B3LYP calcula-
tions performed poorly, consistently underestimating the
barriers. With this settled, the MP2 calculations were
extended to molecules whose activation data were not
accessible.
Elimination reactions of thiosulfinates have been
demonstrated by Block in natural and model com-
pounds.4,5,90,91 A half-life of 29 of 7 min in neat thermoly-
sis at 96 °C91 can be used to verify that the calculations
are at least in the right ballpark. Using the calculated
∆Hq
of 21.3 kcal/mol, a ∆Sq of -14 cal/mol‚K is
elim
obtained from the experimental half-life. If the calculated
value is low by 1 kcal/mol, the newly implied ∆Sq is -11.4
cal/mol‚K.
These calculations demonstrate that substituent effects
on the sulfoxide elimination reaction are much easier to
understand if ∆Hrxn is known in addition to ∆Hq
.
elim
Gr ou p 8: La cton e Elim in a tion s. Finally, the elimi-
nation reactions of the two diastereomers 30 and 31
deserve some comment. Three conformational minima
were found for each diastereomer. As usual, we calculate
the ∆H data by taking the lowest energy conformation
as the starting material.92 The lowest energy conformer
is found for 31, which is 1.6 kcal/mol lower than that of
30.93 These compounds were analyzed to compare to the
work by Trost and co-workers94,95 in which an unspecified
mixture of the two diastereomers yielded mainly the
endocyclic product.
Sometimes it is easier to qualitatively predict the activa-
tion enthalpy of the reverse reaction, that is, the addition
reaction. This is probably an artifact of the ∆Hrxn being
positive in the direction of the elimination. For basically
unactivated olefins and sulfenic acids, ∆Hq
is ap-
addn
proximately 10 kcal/mol. Activation of the sulfenic acid
by making it more nucleophilic results in a lowering of
the barrier. Activation of the olefin by conjugating it to
a carbonyl also lowers the barrier. When the olefin is
transformed into another simple π-bonded system (e.g.,
formaldehyde), the sense of electrophilicity becomes very
important, such that in the “correct” orientation, addition
barriers are virtually nonexistent, whereas they are
higher than usual when the polarity is mismatched. In
the direction of the elimination reaction, many of these
same qualitative predictions on the value of ∆Hq
can
addn
be drawn on the basis of basicity of the sulfinyl group
and the acidity of the abstracted proton as long as ∆Hrxn
is also known. The acidity of the abstracted proton,
however, is a minor issue if comparing structures where
the new π-bond will be conjugated with another π-system
Trost attributed the mixture to the greater stability
of a conformation in which the sulfinyl group and
carbonyl dipoles opposed one another but recognized that
other factors may play a role. We find that the exo
product is slightly favored by about 1 kcal/mol from 30
but that the endo product is formed with a 6.6 kcal/mol
lower barrier from 31 (Table 6). The absolute energy of
31TS-exo stands out as the highest of the four related
Exp er im en ta l Section
Com p u ta tion s. All computations, except the B3LYP,96,97
G2 calculation on methanesulfenic acid and a few semiem-
pirical conformational searches were carried out with the
GAMESS suite of programs.98 Results were visualized with
MacMolPlt.99 The Becke3LYP and G2 calculations were carried
out using GAUSSIAN 94,100 in which the default 6-311 basis
(89) Toutchkine, A.; Aebisher, D.; Clennan, E. L. J . Am. Chem. Soc.
2001, 123, 4966-73.
(90) Block, E. J . Am. Chem. Soc. 1972, 94, 642-4.
(91) Block, E.; O’Connor, J . J . Am. Chem. Soc. 1974, 96, 3929-44.
(92) The barriers between the conformation are of course well below
the energy defference between any of the conformers and either of the
transition states.
(93) The four other conformers are within 5.5 kcal/mol of the lowest
conformer of 31.
(94) Trost, B. M.; Leung, K. K. Tetrahedron Lett. 1975, 48, 4197-
200.
(96) Becke, A. D. Phys. Rev. A 1988, 38, 3098-100.
(97) Lee, C.; Yang, W.; Parr, T. G. Phys. Rev. B 1988, 37, 785-9.
(98) Schmidt, M. W.; Baldridge, K. K.; Boatz, J . A.; Elbert, S. T.;
Gordon, M. S.; J ensen, J . H.; Koseki, S.; Matsunaga, N.; Nguyen, N.;
Su, S. J .; Windus, T. L.; Dupuis, M.; Montgomery, J . A. J . Comput.
Chem. 1993, 14, 1347-63.
(95) Trost, B. M.; Salzmann, T. N.; Hiroi, K. J . Am. Chem. Soc. 1976,
98, 4887-902.
(99) Bode, B. M.; Gordon, M. S. J . Mol. Graphics. Mod. 1998, 16,
133-8.

8736 J . Org. Chem., Vol. 66, No. 26, 2001
Cubbage et al.
set was made to conform with those in GAMESS, as developed
by McLean and Chandler.101 Low energy conformations of
starting materials and products were determined using the
PM3 model, and subsequent optimizations used those confor-
mations as starting geometries. Hessians were obtained to
confirm the nature of the stationary points. The Gonzales-
Schlegel second-order method55 was used for determining
intrinsic reaction coordinate (IRC) paths. Except as noted, the
products (usually an olefin and a sulfenic acid) were calculated
independently and their energies summed. For each molecule,
the coordinates, absolute energy in hartrees, and zero-point
energies are given in the Supporting Information. Activation
barriers and reaction enthalpies are all corrected with zero-
point energies but do not contain any further corrections for
thermal energy at temperatures above 0 K.
the layers were separated. The organic layer was washed with
another portion of aqueous NaOH, dried with MgSO4, and
concentrated in vacuo. Further purification was carried out
as noted. Unoptimized isolated yields in the range of 60-80%
were typical, though near-quantitative yields are sometimes
obtained. Alkyl sulfoxides can be prepared easily at lower
temperature still, such as -78 °C.
Meth yl 3-p h en ylp r op yl su lfoxid e (1e)104 was prepared
by oxidation of the corresponding sulfide:10 1H NMR (CDCl3)
δ 7.33-7.27 (m, 2 H), 7.22-7.18 (m, 3 H) 2.80 (t, J ) 7.5 Hz,
2 H), 2.61-2.73 (m, 2 H), 2.54 (s, 3 H), 2.09-2.17 (m, 2 H);
13C NMR (CDCl3) δ 140.4, 128.6, 128.5, 126.4, 53.8, 38.6, 34.2,
24.2; IR (thin film) 3024, 2922, 2859, 1453, 1044, 747, 700
cm-1
.
Meth yl 2,2,3,3-tetr a d eu ter o-3-p h en ylp r op yl su lfoxid e
(1e-d 4) was prepared in quantitative yield by oxidation of
methyl 2,2,3,3-tetradeutero-3-phenylpropyl sulfide (0.08 g, 0.05
mmol) with m-CPBA as described above.10 It was further
purified by recrystallization from ether at low temperature
yielding white crystals (65% yield): 1H NMR (CDCl3) δ 7.29-
7.26 (m, 2H), 7.21-7.15 (m, 3H), 2.95 (qAB, J ) 13 Hz, ∆ν/J )
5.5, 2H), 2.51 (s, 3H); 13C NMR (CDCl3) δ 140.3, 128.8, 128.5,
126.6, 53.7, 38.6; EI-MS (m/e, relative abundance) 186 (27),
120 (100), 93 (48); IR (thin film) 3021, 2953, 2905, 2206, 2106,
The temperature-dependent KIEs were calculated using the
program ISOEFF98,60 which uses vibrational frequencies from
the substrate and TS to solve for the KIE using Bigeleisen
equation.61,63 The ISOEFF98 program uses Hessian matrixes
obtained from GAMESS output.
Th er m olysis In str u m en ta tion . The stirred-flow reactor
has a temperature-controlled furnace and is modeled very
closely after the one that has been previously described.48 It
uses He as a carrier gas to bring the sample into a quartz
reactor (clean, and silylated) whose volume controls the
residence time, which is a few seconds. Samples were injected
as concentrated solutions in acetonitrile, except where noted.
After the furnace section, the gases are sent to a GC that
operates at lower temperatures, where starting materials and
products are separated and quantified. Rate constants are
extracted from each run, and multiple injections were made
at each temperature. Activation data are extracted from rate
constants obtained over the range of temperatures where both
starting material and product can be quantified. All sulfoxides
thermalized were greater than 99% purity, as determined by
the observation of a single peak by GC without thermolysis.
Com p ou n d P r ep a r a t ion . Gen er a l P r oced u r e. Unless
otherwise noted, starting materials were obtained from Aldrich
and used as received. Characterization was carried out on a
Bruker Avance DXR NMR operating at 400 MHz for proton
and 100 MHz for carbon. The 13C signals for CD2 carbons were
generally not observed due to the low signal-to-noise and high
multiplicity. Mass spectra were obtained on a Finnigan TSQ
700 operating in EI mode. IR spectra were obtained on a
Mattson Galaxy Series FTIR 3000. Dry THF was freshly
distilled from benzophenone ketyl. For the AB quartet in 1a
∆ν/J was calculated using: ∆ν/J ) (4C2 - J 2)1/2/J ; C )
separation from first peak to third peak and J ) separation
from first to second peak in the quartet. Sulfoxides 2e and 3e
were prepared as described in Guo’s dissertation.102 Synthetic
details for other compounds that have appeared in the
literature are in the Supporting Information.
2090, 1301, 1133, 1028, 744, 703 cm-1
.
Meth yl 2,2,3,3,4,4-h exa d eu ter o-3-p h en ylp r op yl su lfox-
id e (1e-d 6) was prepared from methyl 2,2,3,3,4,4-hexadeutero-
3-phenylpropyl sulfide (0.08 g, 0.05 mmol) as described above
in quantitative yield. It was further purified by recrystalliza-
tion from ether at low-temperature yielding white crystals
(65% yield): 1H NMR (CDCl3) δ 7.28-7.24 (m, 2H), 7.19-7.13
(m, 3H), 2.49 (s, 3H); 13C NMR (CDCl3) δ 140.3, 128.6, 128.5,
126.4, 38.5; EI-MS (m/e, relative abundance) 188 (38), 123
(100), 93 (54); IR (thin film) 3021, 2983, 2213, 2111, 1448, 1107,
1040, 744, 703 cm-1. Preparation of the sulfide is described in
the Supporting Information.
3-P h en ylp r op yl Meth a n esu lfin a te (26e).105 A solution of
methanesulfinyl chloride (10.0 g, 0.10 mol)106 in dry ether (80
mL) was added dropwise with stirring and cooling in an ice
bath to a solution of distilled 3-phenylpropan-1-ol (12.6 g, 0.093
mol) and pyridine (8.1 g, 0.10 mol) in ether (20 mL). After the
mixture was stirred overnight, the mixture was poured into
ether (100 mL) and washed with cold water (20 mL), cold HCl
(10%, 20 mL), cold saturated NaHCO3 (20 mL), and cold water,
in that order. The organic layer was dried (MgSO4) and
concentrated via rotary evaporator. The crude product (as
needed in 2 mL to 3 mL aliquots) was purified using flash
chromatography (CH2Cl2) affording a clear liquid: 1H NMR
(CDCl3) δ 7.31-7.26 (m, 2H), 7.21-7.18 (m, 3H), 4.09-3.98
(m, 2H), 2.72 (t, J ) 4.0 Hz, 2 H), 2.064-1.978 (m, 2H), 2.62
(s, 3H); 13C NMR (CDCl3) δ 141.0, 128.5, 128.5, 126.2, 67.5,
44.2, 31.9, 31.9; EI-MS (m/e, relative abundance) 198 (3), 118
(100), 117 (56), 91 (76); IR (thin film) 3025, 2947, 2880, 1603,
Gen er a l P r oced u r e for P r ep a r a tion of Su lfoxid es
fr om Su lfid es. For a good general reference for the oxidation
of sulfides to sulfoxides, see Mata’s review.103 In our hands,
the low-temperature oxidation with 1 equiv of m-CPBA is the
usual preferred method. The following is a general procedure.
To an ice-cooled solution of 2-3 mmol of the sulfide in
methylene chloride (15 mL) was dropwise added a solution of
1.0 equiv of m-CPBA in 25 mL of CH2Cl2. After 2 h, the
mixture was poured in to aqueous NaOH (5%, 50 mL), and
1132, 1017, 907, 744, 701 cm-1
.
Ack n ow led gm en t. We gratefully acknowledge sup-
port of this research from The Research Corporation and
the National Science Foundation (CHE9708327). J .W.C.
thanks the Nelson Foundation and Proctor and Gamble
for support.
Su p p or tin g In for m a tion Ava ila ble: Further computa-
tional details including geometries, NOONs, absolute energies,
and IRCs, along with preparation and spectroscopic charac-
terization of a number of synthetic intermediates and 11e2.
This material is available free of charge via the Internet at
http://pubs.acs.org.
(100) Frisch, M. J .; Trucks, G. W.; Schlegel, H. B.; Gill, P. M. W.;
J ohnson, B. G.; Robb, M. A.; Cheeseman, J . R.; Keith, T. A.; Petersson,
G. A.; Montgomery, J . A.; Raghavachari, K.; Al-Laham, M. A.;
Zakrzewski, V. G.; Ortiz, J . V.; Foresman, J . B.; Cioslowski, J .;
Stevanov, B.; Nanayakkara, A.; Callacombe, M.; Peng, C. Y.; Ayala,
P. Y.; Chen, W.; Wong, M. W.; Andres, J . L.; Replogle, E. S.; Gomperts,
R.; Martin, R. L.; Fox, D. J .; Binkley, J . S.; Defrees, D. J .; Baker, J .;
Stewart, J . P.; Head-Gordon, M.; Gonzalez, C.; Pople, J . A. Gaussian
95; Gaussian, Inc.: Pittsburgh, PA, 1995.
J O0160625
(101) McLean, A. D.; Chandler, G. S. J . Chem. Phys. 1980, 72, 5639-
(104) Baker, R.; Spillet, M. J . J . Chem. Soc. B 1969, 7, 880-3.
(105) Andersen, K. K. J . Org. Chem. 1964, 29, 1953-6.
(106) Drabowicz, J .; Bujnicki, B.; Dudzinski, B. Synth. Commun.
1994, 24, 1207-13.
48.
(102) Guo, Y. Ph.D. Thesis, Iowa State University, 1997.
(103) Mata, E. G. Phosphorus, Sulfur, Silicon 1996, 117, 231-86.